AI is creating new opportunities for public sector leaders to explore evidence, challenge assumptions and make better decisions. The quality of that judgement will depend on knowing where the information comes from, how reliable it is and what might be missing.

Recent government policy developments have brought these questions into sharper focus. On 1 October 2026, the government's Digital Functional Standard, GovS 005, was updated to refer more explicitly to AI and Knowledge and Information Management. This follows May's Data Asset Management Policy, which sets clearer expectations for ownership and quality of critical data across government departments and relevant arm's-length bodies.

Where knowledge gets stuck

The government's State of Digital Government Review, published in January 2025, found that 70% of survey respondents reported a data landscape that was poorly coordinated, lacked interoperability and did not provide a single source of truth. The review drew on perspectives from more than 500 leaders across 120 public organisations.

It's a striking figure, but an understandable one. Organisations are constantly responding to changing priorities, rising demand and tight resources. Teams find workarounds and keep services running, while maintaining shared information, agreeing common definitions and capturing learning get pushed back. Over time, the effects build up, even in organisations with deep data expertise.

• Different teams work from different versions of the same information.
• Data sits in systems that don't easily connect.
• Important context lives in people's heads, and leaves when they do.

AI makes these challenges harder to ignore. It can help people find information faster, but it can also present incomplete or outdated material with convincing certainty. That puts a premium on leaders who can interrogate the evidence and know its limits.

A different conversation around the executive table

The Institute for Government's Whitehall Monitor 2026 found that 6% of senior civil servants belonged to the government digital and data profession in 2025.

That's a measure of profession, not capability. But it raises a question about the other 94%: what do leaders across policy, operations, people, finance and service delivery need to be capable of if their organisations are to become genuinely data-led?

Digital and data specialists bring essential expertise. The wider executive team needs to understand and draw on that expertise, connecting it to decisions and outcomes across the organisation.

Leaders need to know which information matters to their outcomes and how far to trust it. They need to bring expertise together across functions, make decisions when the evidence is incomplete and recognise when the organisation needs to learn more.

Sometimes, that takes the courage to question what appears to be a perfectly reasonable answer, or to admit you don't know.

Think about the conversations this might create. A COO asking whether operational data reflects what frontline teams are experiencing. A CPO considering how much organisational knowledge depends on a few experienced colleagues. A finance director recognising the longer-term value of improving information quality, even when the immediate return is difficult to demonstrate.

AI can help leaders explore these questions, but the judgement remains theirs. Building that judgement across the whole executive team is the capability to develop.

Making time for the work that matters

There's a difficult productivity tension in play: improving data quality, rationalising information, documenting knowledge and connecting systems takes time that organisations under pressure to deliver rarely have.

The cost of leaving it undone is less visible: the time spent searching for information, recreating work that already exists, reconciling conflicting reports or relying on one colleague who knows how something works.

Research into the Productivity J-Curve, by Erik Brynjolfsson and colleagues, helps explain why technologies such as AI require complementary investment in people, processes and organisational capabilities before their full benefits can be realised. More recent MIT Sloan research, published in September 2026, highlights the work that sits behind successful AI experimentation: testing, refining, sharing knowledge and adapting solutions with colleagues.

This work is hard to see, hard to measure and easy to deprioritise when productivity gains are needed now.

The prize is leaders who spend less time establishing which version of the facts to trust, and more time acting on them. Getting there means choosing where to focus, protecting time for the work and recognising its contribution to long-term performance.

When knowledge changes what happens next

This is a pattern found in many organisations, public and private, including those with real data strength. Imagine one with extensive service data. Its dashboards show recurring delays, its customer insight team has identified the frustration, and frontline colleagues know why it keeps happening. Yet the problem persists.

Perhaps policy, operations, finance and digital each own part of the solution, and no one owns the whole.

Perhaps the problem has been documented repeatedly, without anyone changing the underlying system. Perhaps the people who know why don't feel it's their place to say so.

The organisation has information. It has knowledge. It may even have formally captured lessons.

But has it learned?

Three conditions make the difference.

Knowledge moves. People can find and use trusted information across teams. Sharing expertise is part of how work gets done, and leaders bring different perspectives together to solve problems.

Knowledge is challenged. When the figures don't stack up, people feel safe enough to say so, and leaders go first by naming their own blind spots. Leaders invite different views, question assumptions and are prepared to change their minds when new evidence emerges.

Learning changes practice. Someone owns what happens next, insights lead to real change, and the organisation checks whether it helped.

These conditions are shaped by everyday leadership choices: what leaders prioritise, the behaviours they encourage, the conversations they make space for, what they recognise as good performance and whether they look beyond their own function.

Three questions for your leadership team

Start with one important service or decision, and ask:

1. Where are gaps or conflicting versions of information making our decisions harder?
2. When someone questions the evidence or spots something we've missed, how do we respond, and what happens next?
3. What have we actually changed because of something we've learned?

The real test of data-led leadership is what happens when new evidence reaches the executive table. Does it prompt curiosity? Better questions? A different decision? And, ultimately, a change in how the organisation works? How leaders respond to what their organisation knows, and what it still needs to learn, will shape the value they get from AI.

‍